Globally Optimized TDOA High Frequency Source Localization Based on Quasi-Parabolic Ionosphere Modeling and Collaborative Gradient Projection

TL;DR

This paper introduces a novel collaborative gradient projection method combined with particle swarm optimization to improve global localization accuracy of high frequency sources using ionosphere-refracted radio signals, overcoming local optima issues.

Contribution

It develops a PSO-assisted collaborative gradient projection algorithm for globally solving the nonconvex TDOA HF source localization problem based on quasi-parabolic ionosphere modeling.

Findings

The proposed method achieves higher localization accuracy than existing techniques.

Numerical results confirm the effectiveness of PSO-assisted re-initialization in reaching global optima.

The approach demonstrates robustness against practical impairments.

Abstract

We investigate the problem of high frequency (HF) source localization using the time-difference-of-arrival (TDOA) observations of ionosphere-refracted radio rays based on quasi-parabolic (QP) modeling. An unresolved but pertinent issue in such a field is that the existing gradient-type scheme can easily get trapped in local optima for practical use. This will lead to the difficulty in initializing the algorithm and finally degraded positioning performance if the starting point is inappropriately selected. In this paper, we develop a collaborative gradient projection (GP) algorithm in order to globally solve the highly nonconvex QP-based TDOA HF localization problem. The metaheuristic of particle swarm optimization (PSO) is exploited for information sharing among multiple GP models, each of which is guaranteed to work out a critical point solution to the simplified maximum likelihood formulation. Random mutations are incorporated to avoid the early convergence of PSO. Rather than treating the geolocation of HF transmitter as a pure optimization problem, we further provide workarounds for addressing the possible impairments and challenges when the proposed technique is applied in practice. Numerical results demonstrate the effectiveness of our PSO-assisted re-initialization strategy in achieving the global optimality, and the superiority of our method over its competitor in terms of positioning accuracy.